Weber local descriptor with edge detection and double Gabor orientations for finger vein recognition
Lulu Ye, Huabin Wang, Mengli Du, Ying He, Liang Tao · 2019
Difference summation is utilized to calculate the differential excitation of Weber Local Descriptor (WLD), however, this method is sensitive to noise and may offset positive and negative difference; at the same time, the orientation operator of WLD solely calculates the ratio of the gray difference between the vertical and horizontal orientations, that cannot portray the orientation features of the finger vein effectively. According to the features of finger vein images, this paper based on the differential excitation and orientation operator of Weber descriptors, several improvements are as below: 1)firstly, find the edge area in the image, and optimize the gradient magnitude according to the position of the pixel to increase the discrimination, 2)and then differential excitation is represented by the ratio of the optimized gradient magnitude to the current pixel; 3)double Gabor orientations are leveraged to replace the original gradient orientation to reduce the influence of translation and rotation on recognition; 4)finally, in order to better measure the similarity among features, a cross-matching algorithm is used to improve the recognition rate. In this work, the proposed local descriptor and the cross-matching algorithm are combined for finger vein image recognition, the recognition rates reached 100% and 99. 458% on the two finger vein databases at home and abroad, respectively.